Hippocampal band power did not track memory load, but a small spectral-slope effect appeared in 14 patients

We tested whether hippocampal stereo-electroencephalography (SEEG) activity during verbal working-memory maintenance varied with memory load in 14 patients. The analysis used 96 hippocampal bipolar pairs and 1,161 correct, artifact-free trials from sessions 1 and 2 of OpenNeuro ds004752. We fetched seconds 3–6 of each trial from the public S3 copy using HTTP byte ranges; the exact dataset release was not recorded, and SHA-256 hashes identify each fetched chunk. In a 3.5–6.0 s window, band power from theta (4–8 Hz) through high gamma (70–150 Hz) showed no load effect after Holm correction (all adjusted p = 1.0; |mean rho| ≤ 0.041). Response time increased with load (mean rho = +0.23; 13/14 patients; p = 0.0002). The 30–45 Hz aperiodic spectral slope became slightly steeper at higher loads (mean rho = −0.038; SD = 0.043; 12/14 patients negative; p = 0.0052; Holm p = 0.031). In session 1 alone, the slope result was not significant (p = 0.10); session 2 was added after session-1 results were seen. The slope finding is therefore a candidate for testing in sessions 3–8, which were not analyzed here. A weak session-1 high-gamma decrease (rho = −0.067; uncorrected p = 0.043) did not hold in the pooled analysis. Atlas labels were not checked on patient MRI, some bipolar pairs may include contacts outside the hippocampus, many patients had hippocampal sclerosis, and the analysis included no fixation baseline or measures of phase or coupling. The slope fit range was fixed at 30–45 Hz, and only sessions 1–2 of up to 8 sessions per patient were used.

hippocampus · stereo-electroencephalography · working memory · memory load · spectral power · aperiodic slope · verbal Sternberg task · OpenNeuro ds004752